Yes Brabo I already said that we have analysed a lot here, and at the time I have played that game I was unsure that your improvment was better than h4/h6 but now it's clear that it is. And your improvment depends on White's move order. Indeed, as you understand, in a case of an early h4+g4 without a4, Black has to play with h6 entering a difficult endgame but maybe holdable. My game was rated and we played for "e points" in a match with 2 games, one with Black one with White. I have underestimated my opponent, and thought I would either make an easy draw with all my analysis, or suffer to learn something in the queen's ending we analysed here. But instead he crushed me with that g3!! move.

Hello it has been a long time since the last discussion on that subject, I have studied that famous pawns ending again and again...And I have decided to show my last conclusion. I have played a correspondance game in that line as Black, and my first and only try have been dramatic but at the same time very instructive. My opponent has been very impressive by finding novelties in that endgame in a very few days that nobody have seen included myself after so much work. So he should deserves a good part of the merits of my last conclusion.

Here is the game:

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 g6 6.Be3 Bg7 7.f3 O-O 8.Qd2 Nc6 9.Bc4 Nxd4 10.Bxd4 Be6 11.Bb3 Qa5 12.O-O-O Rfc8 13.Kb1 b5 14.Rhe1 Bxb3 15.cxb3 b4 16.Bxf6 bxc3 17.Bxc3 Bxc3 18.bxc3 Rxc3 19.Re3 Rac8 20.Rxc3 Qxc3 21.Qxc3 Rxc3 22.Rc1 Rxc1+ 23.Kxc1 Kf8 24.Kc2 Ke8 25.Kc3 Kd7 26.Kc4 Kc6 27.h4 h6 28.b4 Kb6 29.a4 e6 30.g3!! (a total surprise, an amazing novelty) Ka6 31.f4! Kb6 32.a5+! Ka6 33.g4 Kb7 34.b5 Kc7 35.f5 (Now White is going to force a transposition into the game Rittner-Arnlind which is now considered to be lost for Black)Kd7 36.f6 g5 37.h5 Kc7 38.Kd4 Kd8 39.e5 d5 40.Kc5 Kd7 41.b6 axb6+ 42.Kxb6 Kc8 43.Kc5 Kb7 44.Kd6 d4 45.Ke7 d3 46.Kxf7 d2 47.Kg7 d1Q 48.f7 Qxg4 49.f8Q Qe4 50.Qd6 Qf5 51.Qb6+ Ka8 52.Qd8+ Kb7 53.Qf6 g4 54.Qxf5 exf5 55.e6 g3 56.e7 g2 57.e8Q g1Q+ 58.Kxh6 Qb1 59.Qd7+ Kb8 60.Qd6+ Ka8 61.Qc6+ Kb8 62.Kg5 Qg1+ 63.Kxf5 and this is 1-0 according to the tablebase 6 pieces. This game ending was perfectly played by White.

This guy alone seemed to refute the whole ending with ease. I worked hard to not allow that Rittner-Arnlind transposition but the alternatives were worse. 

27.h4! and 27.b4! are of equal value and should transpose. In my previous analysis I thought that Black was quite ok with ..h6 after h4, where I believed in Black's drawing chances in the queen's endings we have analysed here before. But that game shows that 27.h4! wins against 27..h6 (?!). My improvment for Black now is to play 27..e6! 28.b4! Kb6! 29.a4 and now if 29..h6?! we transpose into my game. So here, we have the last idea of Brabo, which is 29..f6! and I have tried to refute it, but I could not find a winning line for White here. There are many ways to try but still it seems to hold. But now I am very careful because "we never know"... The idea to meet the f4 move with f5 or the g4 move with g5 is very strong. So the last chance for Black to survive is definitively that 29..f6! instead of playing with h6.

But, that's not the end of the story. The lines I have analysed here with h6 are not useless.

27.h4! e6! 28.b4! Kb6! and now instead of 29.a4 f6!, White can try 29.g4!? where 29.. f6? lose to 30.f4! but 29.g4!? should be responded by 29...h6! 30.f4! a6! which has been analysed here and with the conclusion that Black may hold, it's not clear if White is winning but he has good chances.

So Brabo's idea to avoid to play h6 and to play f6 is really good, but cannot be played against every move orders, it only works against an early a4 idea. Against an early h4/h4, Black has to play h6 instead of f6 and has to play the line we have analysed here.

Sorry to be posting quite late. Here is a remark of mine on #19. I'll be back if I find something in all of your interesting postings #20 to #27.

brabo wrote on 03/20/12 at 22:35:25:

This is a nice line but certainly not forced. 33.Ka5 is not best. Instead, 33.g4 is clearly winning. 28. a4 probably does not deserve a question mark.

27... Kb6 28. a4 e6 29. h4 Kc6 30. b5+ Kb6 31. Kb4 f6 32. h5! Kc7! and here my engines say 33. g4! +- 
The winning plan is e.g. 33... Kb6 34. f4 +- with f4-f5 to follow.
Only after say 33... g5 then 34. Ka5 d5 35. exd5 exd5 36. Kb4 d4 37. a5 +-.

There is no need to examine 32. g3 here, I think.

Many thanks ArKheiN and of course also Papageno for your contributions.

I summarize here below the latest status of the analysis.



So now we are where we started. Black is still playing for 2 results: draw or loss. The only difference is that we are 15 moves (or even more) further in the game. I would never voluntary want to play one of the queenendgames out with black but theoretically it could be that it is still a draw as the extra a-pawn gives really very little shelter for the king.

As you can see in the analysis, I added some more lost lines but the mainline was kept. 

Still interested to hear from somebody if improvements on white or blacks side can be made (earlier in the endgame?) as none of both sides can be 100% satisfied with the endconclusion.

ArKheiN wrote on 03/21/12 at 23:40:12:

You are right. I forgot about the triangulations. I was busy till 2AM so quite groggy looking at this fascinating endgame.

A6 probably indeed doesn't work. I don't give up the plan with f6. Maybe it is playable without a6. Besides we also need to look more in detail to g5 as I too easily rejected that move.

A new summary:

To brabo: 28.a4 e6 29.h4 f6? 30.g4! wins for White. For example 30..g5  31.b5+ axb5 32.axb5+ Kb6 33.hxg5 fxg5 34.Kb4 Also strong against an early a6 is to play b5!, if possible in the position after h4/h6.

Good pieces of work, Papageno and Brabo. I had a very tense work of the ending just after the game Giri-Nakamura, using every source I could (the forum and some old threads, the article of Luis Rodi Maletich, the scan of an article by Van der Weide/Bosch from Nic Magazine 98 I got with the help of a chesspub member (Parisestmagique) and a response of that article in another NiC Magazine.) I used my engine with many "handplay" and trial-error.

I wanted to publish some analysis but I didn't because I wanted to keep it a bit secret for a moment. But you are close to my conclusions now. After a lot of work, my conclusion was one day "White is winning", one day "there is a defense for Black" until I have found a new demolition. Today, my conclusion is: the ending is really complicate, but the way most White player plays it is draw. I am not sure if Black can draw against best play, I have found very promising queen's ending for White but I don't know if it's winning or not, too complicated...

I have constructed a big file in chessposition trainer on that ending, but I am not very good to do pgn like you did. The pgn here is just to respond to Papageno.



31..Kc6 is a better try than 31..Kb5 (like you, I have found that 31..Kb5 is losing to h5!!)
After 31..Kc6!, the game is still very hard for Black.

Here are some of the rules I had wrote for myself, and some of them were cited by Brabo and others:

1)Black has to come as quickly as possible to c6, with the King, and it must be via f8.

2) a quick d5 is possible for Black when White push b5 alone.

3) a6 should be play only after White play a4 AND f4. (of course there are exceptions)

4) e6 should be played as fast as possible to not allow a way for White to come on d5 with the King

5) Black should probably play h6 against h4, but have to be really careful to not play ..g5 automatically because of the break with h5+f4 from White. It's not so clear when it's the last moment for Black to play h6, probably when White is able to push the g-pawn to g5. It may be an useful "waiting move" to play h6 automatically after h4.

6) The White idea of f4-f5-f6 followed by e5 with the idea to enter with the King, allowing Black to queen first as Rittner-Arnlind is always dangerous. When White play that idea after h5 gxg5 gxh5, it should be easy draw for Black. When EACH kingside pawns are blocked, that same idea doesn't work for White when in the queenside, White has a4+b4 against a6. because there is no Zugwzwang possible.

7) Another dangerous plan: h4 played followed by g4 and h5 to FORCE Black to take on h5 or White will break with f4. But that idea works mostly with the a pawn still on a2, to let the White king come to a5.

8) When White has played h4/g4/f4/e4 b4/a4, Black can't play without a6, because Rittner-Arnlind wins for White. 

9) When White plays a quick g4 without h4, g5 is generally good for Black

10) the most promising idea for White is to come as fast as possible in c4 with the king, followed by h4!, then g4, then b4 and f4, and to keep the pawn in a2. That idea of keeping the pawn in a2 seems to be totally new ( I have checked every published games of this ending, corr and OTB) and I have found it while I was working on it, and Brabo seems to have found that too.

So, the most dangerous line I have found is what Brabo and Papageno are discussing, the line of the pgn I put. And after 31..Kc6, White should play 32.g5!! (the most promising to me) or 32.Ka4! (still very promising, and Black has to play 32..f6)

My work here continue to queen's endings with a pawn up for White but I am not able to say if it wins or not, but it's the best practical chance for White.

Papageno wrote on 03/20/12 at 14:02:26:

I am neither fully convinced as it is extremely complicated.

However after 27...Kb6 28.a4 e6 29.h4 Kc6 30.b5+ Kb6 31.Kb4 f6 32.h5 Kc7! 33.Ka5 d5! 34.exd5 exd5 35.Kb4 d4 36.Kc4 (On 36.a5 follows Kd6) Kb6 37.h6 d3 38.Kxd3 Ka5 39.f4 Kxa4 40.Kc4 Ka5 41.Kc5 f5 42.g3 Ka4 43.b6 axb6+ 44.Kxb6 and white will eat the pawn on h7 but it is still draw as black can lock the king up with Kf7.

Still there is a mistake in my analysis as after 27...Kb6 28.a4 e6 29.h4 Kc6 30.b5+ Kb6 31.Kb4 f6 32.g3 Kc7 33.Ka5 Kb7 34.h5 works now with same ideas but eliminating d5 as after 34...d5 35.exd5 exd5 36.Kb4 d4 37.a5 black doesn't have Kd6 so loses. Besides black neither has time for 34...Kc7 35.Ka6! , 34...Kb8 35.b6! 34...Ka8 35.Ka6! and all winning for white

So white afterall wins? No as I learned something from the first variation:
27...Kb6 28.a4 e6 29.h4 Kc6 30.b5+ Kb6 31.Kb4 f6 32.g3 Kc7 33.Ka5 d5! (and now it again works as king is on c7 able to chose between the routes b6 and d6 to pick up the queensidepawns) 34.exd5 exd5 35.Kb4 d4 36.h5 d3 37.Kc3 Kb6 38.Kxd3 Ka5 39.h6 Kxa4 40.Kc4 Ka5 41.Kc5 f5 42.f4 and we are back in the earlier mentioned variation. There are still lots of tricks after d5 but my computer thinks black arrives just on time.

I summarize again:


Anyway Papageno many thanks for your contribution as I don't expect many people taking up the challenge to dig in this endgame.

I had the last week, a very extensive look to the famous pawnendgame using many sources.
- Megadatabase games and analysis
- Correspondence games and analysis
- Analysis of chesspub
- Analysis of Argentinian Luis Ernesto Rodi Maletich: http://www.ajedrez.com.ar/rodi_7.htm
- ...

My computerengines were quite helpless in this endgame as often the variations were leading far beyond the horizon of the calculations. 

The endgame is extremely complicated and even after many hours analysis and using all these resources, I don't dare to say that my analysis are bulletproof. I found however plenty of ameliorations on old analysis and many new ideas and tricks.

I believe white has the following mainideas for a win:
- Putting king on a6
- Putting pawn on f6 and then breaking through with the king to pick up the pawn on f7
- Create a free pawn on the queenside and win via triangulations.

Blacks defences are in choosing the right setup on the queenside:
- Blocking with a5 even if this gives a passed b5 pawn for white
- Giving white the free pawn on the queenside but generating sufficient counterplay with d5
- Keep the a-pawn on a7 and organise the defense from the 7the line.

Key I think for both sides, is to keep maximum flexibility. On the other hand if white wants to win, he will need to move forward.

I tried to use a very recent mastergame as tool to explain all the ideas.



White became GM next year. Black became IM next year and won on top Moscow Open: http://www.chessvibes.com/reports/untitled-vladimir-belous-wins-moscow-open/
Chess is a very complex game if even such simple looking positions can lead to so many mistakes.

It was a suprise that Kamsky played a critical line against Dragon (9 0-0-0 12 Bd4), I have seen many games where he plays short castle lines. I will not have time to analyse the games myself tommorow.

Quote:


Ah thats interesting, it makes accepting the draw easier to understand, especially when up against a quick player like Nakamura.

Keano wrote on 01/25/12 at 15:53:36:


I must confess that I thought one of the characteristics of Karpov's style is to react to opening suprises by playing safe and making a short draw, even with white.  Example: game 12 of the 1985 Wch match with the Kasparov gambit in the Taimanov.  Karpov should probably have avoided in in game 16 too...

In any case, I think Giri was well behind on the clock after being surprised by this antique line, which also supports a "safety first" approach.

The more I look at the position, the more I think Black might be better if White tries to press. The only two half open files belong two Black and the double h-pawns can restrain White on the kingside. In other words, Black is more active, even though structurally potentially worse.

Keano wrote on 01/24/12 at 10:51:25:


Of course white can play on, but the problem is that g pawn may became weak. Also, h4 may create weaknesses that blacks king can exploit. But I am no saying that black is better, I think the final position is dead equal.

Thank you MNb. My idea for Black to not allow an early ending is 19..Qe5!?, but I couldn't check if it has been tried. This move seems to give only a minimal advantage for White, Black hope to eat the c-pawn in better conditions, or the h pawn, or to have enough pressure with still the queens on the board...

Let me give it a try. It appears that Van der Tak's analysis is not addressed at the Tal-Portisch game (oh, my memory), but Nedeljkovic-Volpert, Belgrade 1961. This is the mistake of Gufeld though. Portisch played ...a6 while Volpert didn't. The improvement comes at move 31 (in the 16.Nd5 move order at move 29)


This is not complete; some lines continue to move 47. Not all lines are from Van der Tak himself; he refers to Averbach and Slutsky/Shereshevsky.
Then Van der Tak gives the game Rittner-Arndlind, corr BdF 1970. Alas it happens not to be in my database and 70 moves of typing is too much for me, sorry.

I own some probably unknown analysis by Dutch theoretician Van der Tak (published in Schaakschakeringen, the magazine of the Dutch corr. chess federation NBC) that suggested that the famous line 14.Rhe1 Bxb3 15.cxb3 b4 of Tal-Portisch, Oberhausen 1961, should lead to a draw.
The problem remains that Black has hardly winning chances.